I'm not sure I was clear about what annoys me about Stephenson's style. It's not the dialectics, it's that this is the only technique Stephenson ever uses for exposition or backstory.
If he can't figure out how to show you what he wants, he just has one character explain it to another, with whatever petty quibbles he might think of while writing that sequence. The paragraphs might as well all start:
"Of course you know,..."
It's a viable tool, but Stephenson uses it like a sledgehammer, and it's painful to read. A good editor would sort that out quickly, but I think he feels editing is beneath him at this point.
The section of the interview about writing longhand, then typing it in later with minimal changes was a just a confirmation of the problem to me.
Stephenson erudition is forced and artificial, while Eco's is effortless and engaging.
Here's a dissection of Stephenson explaining something:
Character 1: (insert 40 lines of pandering monologue.) Character 2: "I don't think I understand". C1: (50 more lines) C2: "So you're saying (insert leading question)" (repeat.)
What's so sad and tranparent about it is that whoever is designated C1 is an authority on all things in this conversation, while C2 is a moron. In the next chapter, the same pattern applies, but now C2 (the moron) is lecturing someone else!
I first noticed it in the Library-Sumerian History section of Snow Crash, but it's in all his books, especially Quicksilver.
Don't forget the grenades! After killing the guards in a room, finding a large enough chest of grenades would allow you to blow out nearly all of the walls.
Just don't shoot open the chest with grenades in it, or you get the icky white noise explosion sound. Dead, dead, dead.
Grenades were also almost the only way to dust the SS guys (in their bulletproof vests).
Jumpstart profiles are really just high-level installation scripts. They work on nearly every server type without modification. You have to do the legwork once (which I'm sure you've done) to figure out which packages you really want to install. If you're using core Solaris (SUNWCreq), then adding packages, you already have your profile almost written.
I do nearly all of my installs from two profiles. One uses the developer Solaris package (machines which need X and header files), and one uses minimum (+ some essential packages). Sun and I differ on quite what "minimum" really means, though.
These two profiles (+ the JASS package and a few custom finish scripts) are good enough to give me a standard, secure, repeatable system configuration. At this job, I've used them without modification for everything from an Ultra 2 to a SunFire 3800.
The magic is still in the installer, but Jumpstart profiles let you tell it what to do. The hard part is coming up with standard policies for installs, which you seem to already have a book of. Translating that to Jumpstart profiles would save you many hours on *every* install.
Flash Archives may be even better for your purposes, because then you can archive a complete installation as a large install file, and clone it at your leisure. It also reconfigures for you based on the sysidcfg stuff, so you don't have to mess with changing hostnames by hand.
Sun Blueprints publishes an excellent (recent) book on Jumpstart, and most of that is available in the original papers on www.sun.com/blueprints
I wanted to put in a short profile, but the lameness filter choked on the "#" comments. Examples can be found in Blueprints, though.
You have to do one install from CD, or use the pre-installed Solaris on every machine shipped from Sun.
Set up your Jumpstart server on that, create your profiles and finish scripts based on your site policies and requirements. Use that to Jumpstart your *real* Jumpstart server, then recycle the old one. Never install from a stock Solaris CD again.
If you have one host, install from CDs (or DVD). If you have two, grumble and perhaps install from CD. More than that? Set up Jumpstart from the beginning.
When upgrades come along, add them to your Jumpstart server. Mine will install solaris 7, four revisions of Solaris 8, 9 EA (should be removed), and the first release of Solaris 9.
My basic install is: add client to/etc/hosts and/etc/ethers run the "add-client" script to set up netbooting add the profile to "rules", run "check" boot net - install go away for a while return to a fully installed and secured system.
Net booting may not be an option for everyone, though. As I said above, there are alternatives. An air-gapped install network is one method, as are custom CDs.
So you've blasted Sun for not moving into the 21st century, but then admitted that you're simply not allowed to use their next-generation tools? Ouch.
It sounds like an education issue at your place of business (not necessarily *your* education).
Most people hear "Jumpstart" and think "net-boots". That's simply not the whole story. As you may know, no network is required to do a hands-off custom install (still with the confusing name Jumpstart). Burn CD, pop in drive, watch system install and configure itself.
Is Jumpstart as a product banned by name? Is it a corporate policy that all installs must be done by hand? I'd work on reversing *those* decisions, while keeping in mind the valid security concerns (no network traffic during installs, for example).
At any rate Sun is not the source of your problems. They have better installation methods, but you seem to be screwed. Sorry, man.
You're absolutely right, I haven't done one from the commandline in quite some time. Why would I?
You mention that "security" is your goal? What are your security concerns over doing Jumpstart installs? Or LiveUpgrade, for that matter? If you don't want the risk of net booting, you can burn custom JumpStart CDs, which will do the install to *your specifications*, and can be reused. I'd hate to see how you customize packages from the commandline. Are you sure you got that right? Did you leave one out? What about the last server? Is it the same?
For "security" purposes, you should be all over Jumpstart (with or without net boots, depending on the situation). There isn't a better way to ensure correct, repeatable, known configurations.
If you're really concerned about matching packages and patches for a completely known installation, go with Flash Archives. They're faster, too.
I fail to see how automation can possibly reduce security for you, and see quite the reverse. I pity anyone who tries to install a custom Solaris by hand, because no sane person does that. Not even Sun, insane as they may be in other things. They've developed better ways to install their OS, which you won't use because of some perceived "security" hangup. Please explain.
Package management on IRIX is great, but there's nothing wrong with Solaris. Custom installations with Solaris Jumpstart are extremely easy and flexible.
I'm not sure where your number-based package scheme is coming from, because I've not seen any such thing in Solaris (2.5 to 9). You may be confusing patches (which are numbered xxxxxx-yy), with packages (mostly SUNWxxxx).
With the Flash Archives and LiveUpgrade stuff that's been quietly appearing in the later revisions of Solaris 8, it's as good or better than anyone else's installation scheme. Furthermore, you'll have much better control, documentation, and repeatability using automated hands-off Solaris systems. I simply won't do Solaris installs of the CD any more.
Use the tools that are there, they'll make your life much easier.
for (i = length(probs) -1; i > 0; i--){
prod *= probs[i];
prodinv *= 1 - probs[i]; } return prod / (prod + prodinv);
The "apply #'*" multiplies all of the elements of a list (the probabilities) together. The mapcar/lambda stuff subtracts each element from 1 to get the inverse probability, then multiplies those all together.
Enjoy. LISP is fun to read.
(BTW, I reversed that loop since it's easier than trying to get a "less than" sign. Normally I'd count up.)
If g + b is less than 5, the algorithm won't return 0, that would be bogus.
It will return nil, as an indication of failure. That's the purpose of the "unless".
The persistence of Subspace continues to amaze me.
The amount of time that people have spent reverse-engineering and recreating a closed-source game, when there's a staggeringly similar open source project (which predates it by years) is crazy. I haven't seen Subspace in some years, but given that major Subspace sites are still providing the old zipfiles, I doubt it's changed at all. Given the fracturing would go on when a closed-source game changes its protocols, I doubt it *can* change much. Meanwhile xpilot continues to be played and further developed by the community that enjoys it, plus the occasional newcomer.
Perhaps the reason I haven't looked at Subspace in some time is that I don't run Windows (nor does the original question asker), and the reverse-engineers haven't made other platforms a priority.
Never fear, xpilot has you covered. Windows versions are available (with source if you want it), and the game was born and raised on X11.
If you enjoy Subspace, or liked Asteroids, or Thrust (on the C64), you owe it to yourself to check out Xpilot. And if something bugs you about it (like the spartan graphics), get the source and hack away. It's encouraged.
Well, the folks at Bell Labs have a fair bit of experience writing C compilers;-)
Unless they've done major work on it (which I doubt), they didn't really use that knowledge.
The last Plan 9 release was almost apologetic about the C compiler. It worked, but wasn't very advanced, i.e. it would lay out the entire program in flat memory while linking, etc.. A classic case of "good enough".
Alef (their threaded C-like language) was the cool development toy in Plan 9, but it's been dropped from this release.
Is there any real reason why Plan9 is better than modern Unices TODAY?
No. Plan 9 is a really cool system, but it's not the kind of thing you're going to use as your desktop OS. It's stripped down and raw, and is lacking a lot of the functionality we now take for granted. (web browser, etc...)
Plan 9 is very much like V7 UNIX plus some modern tech, and is pretty much an OS-hacker system. It's cool for a research toy, and exploring something different, but no sane person is going to base their network around it.
Consider it a different snapshot of Bell Labs OS work.
According to the articles they provide Plan9 is supposed to be a distributed UNIX, or something.
Or something. Plan 9 is not a UNIX system. It was designed as a separate OS research project, but given the authors, it's natural that some UNIX ideas were used.
It was designed to explore the ideas of distributed computing and the proliferation of commodity hardware.
Maybe i'm missing the point but i get a feeling of "been there - done that".
Yes, but look at the dates on the system. Plan 9 has been there, and was doing that ten years ago.
For some reason, Slashdot folks seem to think Plan 9 is new software. Not so.
Umm... in Unix/Linux/*BSD/etc, is not everything a file?
Not exactly. In Plan 9, everything is a file.
For example, the contents of the current mouse selection is available from a file, whose name escapes me. Also, the distinction between block and character devices is gone from Plan 9.
And just what use would you get from clustered Aqua? That's got nothing to do with the issue.
In some sense, this could be done pretty easily.
1) Port PVM and MPI to Darwin (if not already done). 2) Use bitchin' G4 with OSX as your head system. 3) Send your tasks to the cluster for computation.
Now, there are a few problems, like the fact that PVM and MPI aren't transparent, but require specialized programming in each app, and that you can't get PowerPC nodes as cheaply as recycled x86s, but these are just details, right? Right?
If you went to a company and their spokesperson was rude to you.. Would you EVER use that product ? Would you EVER reccomend that product ?
Depends on the product. If the product suits your needs, then use it. Salesmen lie to me, and I consider that extremely rude. Theo has a short temper, but he and the other OpenBSD developers make a quality product. In many situations, I use it and recommend it. In others I don't. But his attitude has little to do with his product.
I doubt Scott McNealy would be much more forgiving if I wandered into his office and said "What's, uh, the deal with this Solaris thing..." At best, he'll point me to a stack glossy literature...
I still buy shit from Sun.
Don't tell me his personality isn't a determent to the project. I talked with people who are "in" with Free/Net BSD's.. They said FreeBSD and NetBSD could probably actually merge into a common code base but they also said there is little chance in hell of ever merging with OpenBSD due to "personality conflicts"....
Determent how? OpenBSD isn't about marketshare, or making money. Their expansion is based on one thing only: "Is our stuff better than theirs?"
Would one big OpenFreeNetBSDi really be better? Why?
And I actually DID try and find something about ISO images.. A search of their site (at that time) showed nothing.
That's good, but there's still a few more places to check before yelling "Help" on the mailing lists. The archives are one of the best, just to make sure no one asked the same thing yesterday.
Slashdot Mirror
I'm not sure I was clear about what annoys me about Stephenson's style. It's not the dialectics, it's that this is the only technique Stephenson ever uses for exposition or backstory.
If he can't figure out how to show you what he wants, he just has one character explain it to another, with whatever petty quibbles he might think of while writing that sequence. The paragraphs might as well all start:
"Of course you know,
It's a viable tool, but Stephenson uses it like a sledgehammer, and it's painful to read. A good editor would sort that out quickly, but I think he feels editing is beneath him at this point.
The section of the interview about writing longhand, then typing it in later with minimal changes was a just a confirmation of the problem to me.
He'd be a great writer, if he had a good editor.
PRECISELY!
Stephenson erudition is forced and artificial, while Eco's is effortless and engaging.
Here's a dissection of Stephenson explaining something:
Character 1:
(insert 40 lines of pandering monologue.)
Character 2: "I don't think I understand".
C1:
(50 more lines)
C2: "So you're saying (insert leading question)"
(repeat.)
What's so sad and tranparent about it is that whoever is designated C1 is an authority on all things in this conversation, while C2 is a moron. In the next chapter, the same pattern applies, but now C2 (the moron) is lecturing someone else!
I first noticed it in the Library-Sumerian History section of Snow Crash, but it's in all his books, especially Quicksilver.
He just needs a good editor.
Don't even start Quicksilver, then.
Halfway through Cryptonomicon, I thought is was one of the best books I'd read. I didn't feel that way by the end, for other reasons.
The entire time I was reading Quicksilver, I thought "edit, edit, edit".
The book should have been about half as long, and nothing would have been lost.
(there are some good parts in Quicksilver, but the majority is fluff.)
"There could be cubes in there the size of gorillas!"
Marge: I have someone here who can help!
Homer: Batman?
Marge: No, he's a scientist.
Homer: Batman is a scientist!
What do you mean only weapon?
Don't forget the grenades! After killing the guards in a room, finding a large enough chest of grenades would allow you to blow out nearly all of the walls.
Just don't shoot open the chest with grenades in it, or you get the icky white noise explosion sound. Dead, dead, dead.
Grenades were also almost the only way to dust the SS guys (in their bulletproof vests).
Not really.
Jumpstart profiles are really just high-level installation scripts. They work on nearly every server type without modification. You have to do the legwork once (which I'm sure you've done) to figure out which packages you really want to install. If you're using core Solaris (SUNWCreq), then adding packages, you already have your profile almost written.
I do nearly all of my installs from two profiles. One uses the developer Solaris package (machines which need X and header files), and one uses minimum (+ some essential packages). Sun and I differ on quite what "minimum" really means, though.
These two profiles (+ the JASS package and a few custom finish scripts) are good enough to give me a standard, secure, repeatable system configuration. At this job, I've used them without modification for everything from an Ultra 2 to a SunFire 3800.
The magic is still in the installer, but Jumpstart profiles let you tell it what to do. The hard part is coming up with standard policies for installs, which you seem to already have a book of. Translating that to Jumpstart profiles would save you many hours on *every* install.
Flash Archives may be even better for your purposes, because then you can archive a complete installation as a large install file, and clone it at your leisure. It also reconfigures for you based on the sysidcfg stuff, so you don't have to mess with changing hostnames by hand.
Sun Blueprints publishes an excellent (recent) book on Jumpstart, and most of that is available in the original papers on www.sun.com/blueprints
I wanted to put in a short profile, but the lameness filter choked on the "#" comments. Examples can be found in Blueprints, though.
Installations don't have to suck...
You have to do one install from CD, or use the pre-installed Solaris on every machine shipped from Sun.
/etc/hosts and /etc/ethers
Set up your Jumpstart server on that, create your profiles and finish scripts based on your site policies and requirements. Use that to Jumpstart your *real* Jumpstart server, then recycle the old one. Never install from a stock Solaris CD again.
If you have one host, install from CDs (or DVD). If you have two, grumble and perhaps install from CD. More than that? Set up Jumpstart from the beginning.
When upgrades come along, add them to your Jumpstart server. Mine will install solaris 7, four revisions of Solaris 8, 9 EA (should be removed), and the first release of Solaris 9.
My basic install is:
add client to
run the "add-client" script to set up netbooting
add the profile to "rules", run "check"
boot net - install
go away for a while
return to a fully installed and secured system.
Net booting may not be an option for everyone, though. As I said above, there are alternatives. An air-gapped install network is one method, as are custom CDs.
So you've blasted Sun for not moving into the 21st century, but then admitted that you're simply not allowed to use their next-generation tools? Ouch.
It sounds like an education issue at your place of business (not necessarily *your* education).
Most people hear "Jumpstart" and think "net-boots". That's simply not the whole story. As you may know, no network is required to do a hands-off custom install (still with the confusing name Jumpstart). Burn CD, pop in drive, watch system install and configure itself.
Is Jumpstart as a product banned by name? Is it a corporate policy that all installs must be done by hand? I'd work on reversing *those* decisions, while keeping in mind the valid security concerns (no network traffic during installs, for example).
At any rate Sun is not the source of your problems. They have better installation methods, but you seem to be screwed. Sorry, man.
You're absolutely right, I haven't done one from the commandline in quite some time. Why would I?
You mention that "security" is your goal? What are your security concerns over doing Jumpstart installs? Or LiveUpgrade, for that matter? If you don't want the risk of net booting, you can burn custom JumpStart CDs, which will do the install to *your specifications*, and can be reused. I'd hate to see how you customize packages from the commandline. Are you sure you got that right? Did you leave one out? What about the last server? Is it the same?
For "security" purposes, you should be all over Jumpstart (with or without net boots, depending on the situation). There isn't a better way to ensure correct, repeatable, known configurations.
If you're really concerned about matching packages and patches for a completely known installation, go with Flash Archives. They're faster, too.
I fail to see how automation can possibly reduce security for you, and see quite the reverse. I pity anyone who tries to install a custom Solaris by hand, because no sane person does that. Not even Sun, insane as they may be in other things. They've developed better ways to install their OS, which you won't use because of some perceived "security" hangup. Please explain.
Package management on IRIX is great, but there's nothing wrong with Solaris. Custom installations with Solaris Jumpstart are extremely easy and flexible.
I'm not sure where your number-based package scheme is coming from, because I've not seen any such thing in Solaris (2.5 to 9). You may be confusing patches (which are numbered xxxxxx-yy), with packages (mostly SUNWxxxx).
With the Flash Archives and LiveUpgrade stuff that's been quietly appearing in the later revisions of Solaris 8, it's as good or better than anyone else's installation scheme. Furthermore, you'll have much better control, documentation, and repeatability using automated hands-off Solaris systems. I simply won't do Solaris installs of the CD any more.
Use the tools that are there, they'll make your life much easier.
It goes something like this:
prod = probs[0];
prodinv = 1 - probs[0];
for (i = length(probs) -1; i > 0; i--){
prod *= probs[i];
prodinv *= 1 - probs[i];
}
return prod / (prod + prodinv);
The "apply #'*" multiplies all of the elements of a list (the probabilities) together.
The mapcar/lambda stuff subtracts each element from 1 to get the inverse probability, then multiplies those all together.
Enjoy. LISP is fun to read.
(BTW, I reversed that loop since it's easier than trying to get a "less than" sign. Normally I'd count up.)
One minor nit.
If g + b is less than 5, the algorithm won't return 0, that would be bogus.
It will return nil, as an indication of failure. That's the purpose of the "unless".
Remember, 0 isn't false in LISP.
The persistence of Subspace continues to amaze me.
The amount of time that people have spent reverse-engineering and recreating a closed-source game, when there's a staggeringly similar open source project (which predates it by years) is crazy.
I haven't seen Subspace in some years, but given that major Subspace sites are still providing the old zipfiles, I doubt it's changed at all. Given the fracturing would go on when a closed-source game changes its protocols, I doubt it *can* change much. Meanwhile xpilot continues to be played and further developed by the community that enjoys it, plus the occasional newcomer.
Perhaps the reason I haven't looked at Subspace in some time is that I don't run Windows (nor does the original question asker), and the reverse-engineers haven't made other platforms a priority.
Never fear, xpilot has you covered. Windows versions are available (with source if you want it), and the game was born and raised on X11.
If you enjoy Subspace, or liked Asteroids, or Thrust (on the C64), you owe it to yourself to check out Xpilot. And if something bugs you about it (like the spartan graphics), get the source and hack away. It's encouraged.
The main page is www.xpilot.org, and a fine beginner's guide with Windows binaries can be found at www.j-a-r-n-o.nl/xpilot.html.
Enjoy.
first post baby!
Eat it, SUCKERS!
Gestures are a nice thing to have in a window manager.
There was a patch to let FVWM2 use gestures, and I once modified wmx to use the same guy's library.
Pester your friendly neighborhood window manager team for gesture support, it's fun.
If anyone actually cares, and uses wmx, I probably still have the patch.
Yeah, but can you create new net connections by writing to /proc/net/tcp?
Don't be so defensive about UNIX, it doesn't have every good idea...
Just ask the guys who wrote it...
Well, the folks at Bell Labs have a fair bit of experience writing C compilers ;-)
Unless they've done major work on it (which I doubt), they didn't really use that knowledge.
The last Plan 9 release was almost apologetic about the C compiler. It worked, but wasn't very advanced, i.e. it would lay out the entire program in flat memory while linking, etc.. A classic case of "good enough".
Alef (their threaded C-like language) was the cool development toy in Plan 9, but it's been dropped from this release.
You might think that this is just a troll, but it's not far off.
Plan 9 rocks.
Is there any real reason why Plan9 is better than modern Unices TODAY?
No. Plan 9 is a really cool system, but it's not the kind of thing you're going to use as your desktop OS. It's stripped down and raw, and is lacking a lot of the functionality we now take for granted. (web browser, etc...)
Plan 9 is very much like V7 UNIX plus some modern tech, and is pretty much an OS-hacker system. It's cool for a research toy, and exploring something different, but no sane person is going to base their network around it.
Consider it a different snapshot of Bell Labs OS work.
According to the articles they provide Plan9 is supposed to be a distributed UNIX, or something.
Or something. Plan 9 is not a UNIX system. It was designed as a separate OS research project, but given the authors, it's natural that some UNIX ideas were used.
It was designed to explore the ideas of distributed computing and the proliferation of commodity hardware.
Maybe i'm missing the point but i get a feeling of "been there - done that".
Yes, but look at the dates on the system. Plan 9 has been there, and was doing that ten years ago.
For some reason, Slashdot folks seem to think Plan 9 is new software. Not so.
Umm... in Unix/Linux/*BSD/etc, is not everything a file?
Not exactly. In Plan 9, everything is a file.
For example, the contents of the current mouse selection is available from a file, whose name escapes me.
Also, the distinction between block and character devices is gone from Plan 9.
And just what use would you get from clustered Aqua? That's got nothing to do with the issue.
In some sense, this could be done pretty easily.
1) Port PVM and MPI to Darwin (if not already done).
2) Use bitchin' G4 with OSX as your head system.
3) Send your tasks to the cluster for computation.
Now, there are a few problems, like the fact that PVM and MPI aren't transparent, but require specialized programming in each app, and that you can't get PowerPC nodes as cheaply as recycled x86s, but these are just details, right? Right?
Rock on, Mac Beowulf Darwinthing!
Is it childish ? A little. Tough.. 1st impressions are REALLY important.
You're absolutely right.
Consider the first impression you've made: "I don't really feel like looking stuff up, I'll just pester someone else..."
If you went to a company and their spokesperson was rude to you.. Would you EVER use that product ? Would you EVER reccomend that product ?
Depends on the product. If the product suits your needs, then use it. Salesmen lie to me, and I consider that extremely rude. Theo has a short temper, but he and the other OpenBSD developers make a quality product. In many situations, I use it and recommend it. In others I don't. But his attitude has little to do with his product.
I doubt Scott McNealy would be much more forgiving if I wandered into his office and said "What's, uh, the deal with this Solaris thing..." At best, he'll point me to a stack glossy literature...
I still buy shit from Sun.
Don't tell me his personality isn't a determent to the project. I talked with people who are "in" with Free/Net BSD's.. They said FreeBSD and NetBSD could probably actually merge into a common code base but they also said there is little chance in hell of ever merging with OpenBSD due to "personality conflicts"....
Determent how? OpenBSD isn't about marketshare, or making money. Their expansion is based on one thing only: "Is our stuff better than theirs?"
Would one big OpenFreeNetBSDi really be better? Why?
And I actually DID try and find something about ISO images.. A search of their site (at that time) showed nothing.
That's good, but there's still a few more places to check before yelling "Help" on the mailing lists. The archives are one of the best, just to make sure no one asked the same thing yesterday.